About ClearGrid
ClearGrid is on a mission to
revolutionize the debt resolution industry. We’re using
AI, automation, and real-time data to completely rethink how debt collection works, unlocking better outcomes for institutions and individuals alike. Think of us as rebuilding the collections —starting from scratch, with modern tools.
We're a fast-growing startup with ambitious goals, and we’re looking for builders who are hungry to make a real dent in a space with massive untapped potential.
About The Role
This is a hands-on, technical role that sits at the intersection of AI, language, and financial services. You will design and own AI agents that handle real customer conversations in Arabic and English, operating under strict compliance requirements set by our lender partners. The role demands equal parts linguistic precision, technical understanding of how LLMs work, and deep familiarity with financial services norms.
What You’ll Do
Agent Design & Prompt Engineering
- Write and own system instructions for AI voice agents end-to-end: persona definition, tone calibration, identity verification scripts, and dialect-specific behaviour for Gulf Arabic audiences.
- Manage agents across the full customer contact lifecycle, from early-stage reminder agents through to payment arrangement, broken promise, and compliance-focused agents.
- Encode lender-specific rules directly into system instructions.
- Inject dynamic customer and account variables provided by the data team to personalise each agent interaction.
Daily Testing & Quality Assurance
- Run daily test calls to address a wide range of real-world scenarios.
- Listen to live and recorded calls to identify bugs, off-script responses, or compliance gaps, and ship hotfixes quickly.
- Work within a structured QA feedback loop, acting on review comments to improve agent quality before deployment.
- Coordinate with cross-functional teams to align on call triggers, variable feeds, and product rules.
Voice & Speech Optimisation
- Develop new agent voices and evaluate them for naturalness, clarity, and suitability per lender and dialect.
- Tune STT/ASR (speech-to-text), TTS (text-to-speech), and latency parameters to maintain call quality.
- Control temperature and model parameters to keep agents within compliant, predictable behaviour, preventing hallucination and manipulation attempts.
Continuous Improvement
- Monitor developments in AI tooling, voice technology, and LLM capabilities and assess their relevance to the agent stack.
- Study real customer call dynamics alongside the operations team and use those insights to improve agent realism and effectiveness.
- Maintain detailed documentation for each prompt version across all client projects.
Required
What We’re Looking For
- Native or near-native Arabic speakers.
- Demonstrated experience in AI prompt engineering or training AI/LLM models in a professional setting.
- Working knowledge of voice AI stack: STT/ASR, TTS, latency management, hallucination mitigation.
- Understanding of how LLMs think - temperature, constraints, failure modes, and manipulation resistance.
- Experience or background in banking, debt collection, fintech, or financial services.
- Ability to produce structured, unambiguous written prompts in both English and Arabic.
- High attention to detail and comfort with repetitive testing and incremental iteration.
Nice to Have
- Hands-on experience with voice cloning tools and multi-voice agent pipelines.
- Familiarity with debt collection workflows: bucket segmentation, PTP cycles, RPC scripts, and lender compliance requirements.
- Prior exposure to Gulf-based financial institutions or BNPL platforms.
- Experience coordinating with QA teams and managing structured feedback loops.
